Comparative evaluation of techniques for missing rainfall data estimation in arid regions: case study of Al-Madinah Al-Munawarah, Saudi Arabia

Rainfall plays an essential part in numerous aspects of the natural world, including the environment, ecosystems, human societies, and the global climate system. The lack of rainfall data, typically accompanied by data gaps in arid regions like Saudi Arabia, presents substantial obstacles for hydrological and environmental studies. The objective of this study is to identify a suitable imputation technique for finding missing rainfall data on daily and monthly time scales. In this study, eight weather stations were selected, located in the vicinity of Al-Madinah Al-Munawarah City for the period of 5 years (2008–2012). Two stations (226, 371) were considered target/empty stations to compare the computed values from each method to real values, i.e., cross-validation. In this study, various techniques, including arithmetic average (AA), inverse distance weighing (IDW), normal ratio (NR), satellite products, TRMM, IMERG-GPM, CHIRPS, MERRA-2, and artificial intelligence-based and feed-forward backpropagation neural network (FFBP-NN), were evaluated. Statistical measures were used to check the reliability of each imputation technique on daily and monthly rainfall datasets. The results revealed that FFBP-NN exhibited the highest correlation values, surpassing 0.95 for both the stations on monthly and above 0.80 on daily time scale. IMERG-GPM performed well across satellite datasets, with a daily correlation over 0.50 and a monthly correlation above 0.80. Similarly, NR outperformed AA and IDW techniques in terms of correlation, providing values above 0.5 for daily and 0.89 for monthly intervals over both stations. Generally, all methods performed well on both time scales, except MERRA-2 dataset having a lower correlation coefficient. Based on the analysis, it is recommended to utilize the FFBP-NN approach for longer time series data availability while IMERG-GPM for high spatial variation in region. This research contributes to the ongoing efforts to mitigate data gaps in arid regions and supports more accurate water resource management and environmental planning.